AI in recruitment: start small. But start deliberately.

For many teams, AI in recruitment feels like a choice between two bad options.

Either do nothing, because GDPR, the EU AI Act and the risk of treating candidates unfairly feel overwhelming. Or automate too much, too quickly, because suddenly everyone has to be “doing something with AI.”

There is a better third option: start small, clearly bounded and human-led.

Not every AI application should make a decision

Under the EU AI Act, certain AI systems used for recruitment and selection sit in a high-risk category. That includes systems intended to analyse, filter or evaluate candidates. That does not mean AI has no place in recruitment. It does mean it is not a shiny object to drop into your hiring process without thinking.

GDPR also draws an important line. People generally have the right not to be subject to a decision based solely on automated processing when it produces legal or similarly significant effects. In a selection context, that is a good reason to treat human review as an essential part of the process, not a box to tick.

Those are not legal opinions, and they are not two rules that magically make a process “compliant.” They are a useful reality check: technology should make your process better, not more opaque.

Start with a question that helps a human do better

The most valuable early AI use cases are often not the ones where a tool decides who moves forward. They are the applications that help a recruiter have better conversations and look more closely at observable behaviour.

Think of a bounded step that adds another layer of observation without allowing a score to have the final word. How does someone approach an unfamiliar problem? What choices do they make when information is incomplete? Do they adapt their approach when the context changes?

Signals like these do not deliver a definitive verdict on someone’s potential. But they do give you material for a richer conversation.

Human judgment is not a checkbox

A human in the loop is not a person who clicks “approve” after the system has already decided. Meaningful human involvement means someone:

• understands what the tool does and does not measure: a score is a starting point for interpretation, not a personality diagnosis;
• can intervene: a recruiter or hiring manager must be able to question, qualify or set aside an output;
• knows the context: a candidate is more than one answer, one CV or one data point;
• can explain what happened: candidates deserve a clear conversation about how the process works and the role technology plays in it.

That matches the direction of the EU AI Act. For high-risk systems, Article 26 includes duties around following instructions for use, assigning oversight to competent people and monitoring operation. The exact duties depend on the use case and your organisation’s role. But the underlying question is surprisingly simple: who is actually keeping their hands on the wheel?

AI because it is better, not because it is on your roadmap

At Assess4me, we do not use game-based assessments to automate the human decision away. We use them to help recruiters observe behaviour in action.

In a game context, you see more than what someone says about themselves. You gain a prompt to discuss choices, problem-solving, attention, collaboration and adaptability. The recruiter retains judgment. The assessment adds another behavioural lens.

For us, that is the mature first step with AI in recruitment: small enough to understand, clear enough to discuss and human enough to take responsibility for.

You do not need to sprinkle AI everywhere to make progress. Sometimes the smartest automation is the one that helps you look at people more carefully again.

Sources: Regulation (EU) 2024/1689, Annex III and Article 26; GDPR, Article 22. This article provides general information, not legal advice.